Identification of existing pharmaceuticals and herbalmedicines as inhibitors of SARS-CoV-2 infectionJia-Tsrong Jana, Ting-Jen Rachel Chenga

, Yu-Pu Juangb, Hsiu-Hua Maa, Ying-Ta Wua

, Wen-Bin Yanga,Cheng-Wei Chenga

, Xiaorui Chena, Ting-Hung Chouc

, Jiun-Jie Shiec, Wei-Chieh Chenga, Rong-Jie Cheinc,

Shi-Shan Maoa, Pi-Hui Lianga,b,1, Che Maa,1, Shang-Cheng Hunga,1

, and Chi-Huey Wonga,d,1

aGenomics Research Center, Academia Sinica, Taipei 115, Taiwan; bSchool of Pharmacy, National Taiwan University, Taipei 110, Taiwan; cInstitute ofChemistry, Academia Sinica, Taipei 128, Taiwan; and dDepartment of Chemistry, The Scripps Research Institute, La Jolla, CA 92037

Contributed by Chi-Huey Wong, December 11, 2020 (sent for review October 16, 2020; reviewed by Michael D. Burkart, Kuo-Hsiung Lee,and Yasuhiro Kajihara)

The outbreak of COVID-19 caused by SARS-CoV-2 has resulted inmore than 50 million confirmed cases and over 1 million deathsworldwide as of November 2020. Currently, there are no effectiveantivirals approved by the Food and Drug Administration to containthis pandemic except the antiviral agent remdesivir. In addition, thetrimeric spike protein on the viral surface is highly glycosylated andalmost 200,000 variants with mutations at more than 1,000 positionsin its 1,273 amino acid sequence were reported, posing a major chal-lenge in the development of antibodies and vaccines. It is thereforeurgently needed to have alternative and timely treatments for thedisease. In this study, we used a cell-based infection assay to screenmore than 3,000 agents used in humans and animals, including 2,855small molecules and 190 traditional herbal medicines, and identified15 active small molecules in concentrations ranging from 0.1 nM to50 μM. Two enzymatic assays, along with molecular modeling, werethen developed to confirm those targeting the virus 3CL proteaseand the RNA-dependent RNA polymerase. Several water extracts ofherbal medicines were active in the cell-based assay and could befurther developed as plant-derived anti–SARS-CoV-2 agents. Some ofthe active compounds identified in the screen were further testedin vivo, and it was found that mefloquine, nelfinavir, and extracts ofGanoderma lucidum (RF3), Perilla frutescens, and Mentha haplocalyxwere effective in a challenge study using hamsters as disease model.

SARS-CoV-2 | drug repurposing | antiviral | cell-based and animal studies

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is an enveloped, positive-sense, single-stranded RNA

coronavirus of the betacoronaviridae family (1), and the pathogenis responsible for the global pandemic that causes the coronavirus-induced disease in 2019 (COVID-19). Compared to the SARS-CoVand Middle East respiratory syndrome coronavirus (MERS-CoV)outbreaks in 2002 and 2012, respectively, SARS-CoV-2 shows alower fatality rate, but a much higher transmission rate, causing agreater threat to the public health and extraordinary social andeconomic burdens (2).Infection of SARS-CoV-2 starts with the interaction of tri-

meric viral spike (S) protein with human angiotensin-convertingenzyme 2 (ACE2) receptor on airway epithelial cells, followed byviral entry and priming of human transmembrane protease serine2 (TMPRSS2) that cleaves the S protein and initiates viral fusion(SI Appendix, Fig. S1) (3). After entry, the viral genomic RNA istranslated to polyprotein 1a (PP1a) and polyprotein 1ab(PP1ab), which are subsequently cleaved by a papain-like (PL)protease and a 3C-like (3CL) protease to form 16 nonstructuralproteins (Nsp1-16) as a replication-transcription complex. Fourstructural proteins (spike, envelope, membrane, and nucleocap-sid) are encoded at the 3′ end and play important roles in virusmaturation and infection. Replication of viral RNA from the N toC termini of PP1ab is accomplished by replication-transcriptioncomplex proteins, such as RNA-dependent RNA polymerase(Rdrp, Nsp12). The viral proteins further undergo posttransla-tional modifications (such as glycosylation) at the endoplasmic

reticulum (ER)-Golgi intermediate compartment, after whichthey are transported to the cell membrane for exocytosis (4).To date, the clinical management of COVID-19 is mostly

based on supportive care, although several agents targeting viralreplication and inflammation have been reported (SI Appendix,Fig. S1). Remdesivir, an Rdrp prodrug inhibitor, is the onlyantiviral agent approved by the US Food and Drug Adminis-tration for the treatment of COVID-19 (5). Favipiravir, an in-hibitor of influenza Rdrp, was used for the treatment of COVID-19 in Russia, China, and India, but patients receiving the drugmust be closely monitored to prevent adverse events; recently,the result of a phase 3 trial in Japan showed some positive effect.Hydroxychloroquine, especially in combination with a zinc sup-plement, has been reported to exhibit antiviral activity againstRNA viruses, but the clinical use of hydroxychloroquine alonefor the treatment of COVID-19 was halted due to a lack ofsignificant benefit (6).It is well known that RNA viruses have higher mutation rates

than DNA viruses. Recently, a protein interaction map revealed332 human proteins interacting with 27 SARS-CoV-2 proteins(7), and a phosphoproteomic approach was further employed to

Significance

COVID-19 is a global pandemic currently lacking an effectivecure. We used a cell-based infection assay to screen more than3,000 agents used in humans and animals and identified 15with antiinfective activity, ranging from 0.1 nM to 50 μM. Wethen used in vitro enzymatic assays combined with computermodeling to confirm the activity of those against the viralprotease and RNA polymerase. In addition, several herbalmedicines were found active in the cell-based infection assay.To further evaluate the efficacy of these promising compoundsin animal models, we developed a challenge assay with ham-sters and found that mefloquine, nelfinavir, and extracts ofGanoderma lucidum (RF3), Perilla frutescens, and Menthahaplocalyx were effective against SARS-CoV-2 infection.

Author contributions: C.M., S.-C.H., and C.-H.W. designed research; J.-T.J., T.-J.R.C., H.-H.M.,Y.-T.W., C.M., and S.-C.H. performed research; T.-J.R.C., W.-B.Y., T.-H.C., J.-J.S., W.-C.C., R.-J.C.,and S.-S.M. contributed new reagents/analytic tools; J.-T.J. and T.-J.R.C. performed assays;Y.-T.W. did themodeling; C.M. analyzed protein data; J.-T.J., T.-J.R.C., Y.-P.J., Y.-T.W., C.-W.C.,X.C., P.-H.L., C.M., S.-C.H., and C.-H.W. analyzed data; and Y.-P.J., P.-H.L., S.-C.H., and C.-H.W.wrote the paper.

Reviewers: M.D.B., University of California, San Diego; K.-H.L., University of North Carolina atChapel Hill; and Y.K., Osaka University.

The authors declare no competing interest.

This open access article is distributed under Creative Commons Attribution License 4.0(CC BY).1To whom correspondence may be addressed. Email: phliang@ntu.edu.tw, cma@gate.sinica.edu.tw, schung@gate.sinica.edu.tw, or wong@scripps.edu.

This article contains supporting information online at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2021579118/-/DCSupplemental.

Published January 15, 2021.

PNAS 2021 Vol. 118 No. 5 e2021579118 https://doi.org/10.1073/pnas.2021579118 | 1 of 8

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

expand the study of viral–host interaction (8). However, theproteomic analysis reported recently was only focused on the Sprotein and the detailed functions of glycosylation remainedunclear (9). Nevertheless, the S protein is a promising target fordevelopment of neutralizing antibodies and vaccines due to itsexpression on the viral surface and its involvement in host cellentry (10). The S protein is highly glycosylated and broadlymutated, with ∼90% of the sequence being changed, indicatingthe challenge in the development of effective vaccines or anti-bodies with broadly protective activities and the need to developalternative therapies (11). However, development of new ther-apeutics often takes years; therefore, repurposing or reposi-tioning of existing pharmaceuticals and herbal medicines for thetreatment of COVID-19 has been considered as an attractiveapproach.In this study, we screened a library of 2,855 drugs approved for

the treatment of human and animal diseases, as well as 190supplements and traditional Chinese herbal medicines to identifythe inhibitors of SARS-CoV-2 infection to Vero E6 cells. Theeffective compounds identified from the screening were furtherstudied to establish the dose–response relationship. The com-pounds that target the proteolytic process, RNA replication, andglycosylation were collected and further evaluated by a targetenzyme assay and computer simulation to generate a betterunderstanding of their mode of action. Several active compoundsand herbal extracts identified from the cell-based and enzyme-based assays were further evaluated in vivo for their antiinfectiveeffects in hamsters infected with the virus. This investigationidentified several promising candidates with potential for furtherdevelopment.

Results and DiscussionsCell-Based Screening. Initially, the antiviral activity of compoundswas assessed as previously described (12) by visualization of theextent of cytopathogenic effect (CPE) on Vero E6 cells wheninfected with a strain of SARS-CoV-2 from Taiwan Centers forDisease Control (hCoV-19/Taiwan/4/2020, isolated from thethroat swab of a confirmed 39-y-old male patient from Taiwan)at concentrations of 10 μM, 3.3 μM, and 1 μM, respectively (or inthe range from 1 nM to 100 nM for potent compounds). Vero E6cells are African green monkey kidney epithelial cells that arestable cell lines expressing a high level of the ACE2 receptor andhave been used for SARS-CoV research extensively since 2003(12). In recent study, they have been used in evaluating SARS-CoV-2infection and replication by measuring viral-induced CPE (3, 13). Ofthe 2,855 compounds tested, 15 were found to exhibit antiinfectiveeffects in Vero E6 cells and their structures are shown in Fig. 1. Theactivities of these compounds were evaluated on days 3 and 5, andthe minimal concentrations that showed antiinfective effect wererecorded (Table 1). The concentrations of an agent required to in-hibit 50% (IC50) of virus replication and its 50% cytotoxicity (CC50)are shown in Fig. 2 and Table 1.

Protease Inhibitors. To understand the mode of action, some ac-tive compounds were subjected to target-based assay, and thetwo major proteases of SARS-CoV-2, PL protease and 3CL mainprotease, were chosen as targets. When tested at 100 μM, dro-nedarone (12) showed a significant inhibition against PL prote-ase, while the other two known protease inhibitors [nelfinavirmesylate (1) and boceprevir (2)] showed inhibition against the3CL protease (SI Appendix, Table S1). The inhibition of sub-strate cleavage by the enzyme was further confirmed by theHPLC analysis of cleaved substrate (SI Appendix, Fig. S5).Nelfinavir mesylate (1) is an inhibitor of HIV protease with

proven efficacy as an inhibitor of SARS-CoV replication (14). Inour assay, nelfinavir mesylate (1) showed inhibition of SARS-CoV-2replication with IC50 of 3.3 μM (Table 1). Although the clinical resultof lopinavir and ritonavir combination therapy showed no significant

benefits and lowered the anticipation of drug repurposing with HIVprotease inhibitors (15, 16), nelfinavir was found in this study to beactive in the cell-based and target-based assays and is one of the fewcompounds achieving higher plasma concentration than the reportedones in the normal dosing interval (17). The library used in thescreen also contains several other clinically approved HIV proteaseinhibitors, including indinavir sulfate, saquinavir mesylate, atazanavir,ritonavir, darunavir, amprenavir, and lopinavir, and the hepatitis Cvirus (HCV) protease inhibitors daclatasvir, danoprevir and telap-revir, but none of these showed inhibition against SARS-CoV-2 inthe concentrations used in the screen, except boceprevir (2), whichdisplayed a relatively weak activity (IC50 of 50.1 μM) (Fig. 2 andTable 1).The inhibition of nelfinavir mesylate and boceprevir, along

with other HIV protease inhibitors, against the SARS-CoV-2

Fig. 1. Representative drugs showed antiinfective effects at 10 μM. Thesedrugs are categorized according to their potential mode of action againstSARS-CoV-2.

2 of 8 | PNAS Jan et al.https://doi.org/10.1073/pnas.2021579118 Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-

2 infection

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

3CL protease (nsp 5) was measured in vitro subsequently todetermine the Ki values, and the result showed that boceprevir(Ki = 4.8 μM) (Fig. 3 and SI Appendix, Fig. S5) is more potentthan nelfinavir mesylate (Ki = 38.8 μM). We also tested JJS-0309,a TL-3 derivative active against the protease of feline immuno-deficiency virus, HIV, and SARS-CoV (18), but it did not showany inhibition against SARS-CoV-2 in the cell-based assay,

indicating the differences between the two SARS proteases(Fig. 3).The inconsistency between the Ki value and the antiviral ef-

ficacy in the cell-based assay for nelfinavir mesylate and boce-previr prompted us to evaluate their target binding withcomputer modeling. The result of molecular docking showedthat the interactions between nelfinavir mesylate (Ki of 38.8 μM)

Table 1. In vitro anti–SARS-CoV-2 assay

Name Drug class Drug indication

CPE (μM)*

IC50 (μM)† CC50 (μM)‡D3 D5

Nelfinavir 1 Antiviral agent Anti-HIV infection 5.0 10.0 3.3 12.3Boceprevir 2 Antiviral agent Anti-HCV infection 10.0 10.0 50.1 >10Thioguanine 3 Antineoplastic agent Anticancer 1.25 2.5 1.7 25.4Cepharanthine 4 Antineoplastic agent Anticancer 3.75 7.5 2.8 12.9Emetine 5 Antiprotozoal Antiamoebic 5.0 10.0 4.0e-4 >10Ivermectin 6 Anthelmintic Antiparasitic 2.5 —

§ 4.1 13.2Moxidectin 7 Anthelmintic Antionchoceriasis infection 5.0 10.0 3.1 6.9Mefloquine 8 Antimalaria Prevention and treatment of malaria 5.0 10.0 3.2 >10Ivacaftor 9 CFTR potentiators Cystic fibrosis 5.0 N.I. 3.7 12.9Azelnidipine 10 Calcium channel blocker Antihypertension 5.0 N.I. 5.3 12.9Penfluridol 11 First generation antipsychotic Schizophrenia 5.0 10.0 2.4 12.9Dronedarone 12 Ion channel blocker Cardiac arrhythmia 7.5 N.I. 4.5 12.1Salinomycin 13 Polyether antibiotic Prevent coccidiosis of animals 0.039 0.156 4.8e-4 13.1Monensin 14 Polyether antibiotic Prevent coccidiosis of animals 0.117 2.5 6.4 6.6Maduramicin 15 Polyether antibiotic Prevent coccidiosis of animals 0.039 0.117 1.3 3.4

CTFR, cystic fibrosis transmembrane conductance regulator; N.I., no inhibition.*The minimal dose that showed antiinfective effect after virus challenge after incubation for 72 h (D3) and 120 h (D5).†100 PFU of SARS-CoV-2 (multiplicity of infection 0.025) in Vero E6 cells.‡Vero E6 cells.§Cytotoxicity was observed.

Fig. 2. Dose–response relationships of 15 selected antiviral compounds. Vero E6 cells were pretreated with compounds at indicated doses followed bySARS-CoV-2 infection for 48 h. The percentage of viral titer determined by antinucleocapsid antibody after drug treatment (red) and cell viability (blue) weremeasured and expressed as mean ± SD of at least three independent experiments.

Jan et al. PNAS | 3 of 8Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-2infection

https://doi.org/10.1073/pnas.2021579118

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

and SARS-CoV-2 3CL protease were mainly through residuesGly143, Glu166, and Gln189 (Fig. 4A), and as for boceprevir (Kiof 4.8 μM), more interactions were observed with His41, Gly143,His164, and Glu166 residues (Fig. 4B), consistent with the ob-served higher inhibition activity in the enzyme assay. However,nelfinavir mesylate exhibited 15-fold higher anti–SARS-CoV-2activity than boceprevir in the cell-based assay, probably due toits inhibition of multiple targets or the differences in cell per-meability or in cell fusion caused by the S protein (19).

Guanine Analog. Thioguanine (3), an antimetabolite used for thetreatment of cancers and autoimmune diseases, showed an IC50of 1.7 μM. Since it is an analog of guanine, it is expected to act asa prodrug and after conversion into thioguanosine triphosphate(TGTP), it could act as an inhibitor of Rdrp or GTP-bindingprotein associated with cellular DNA synthesis and replication(20). The molecular docking also indicated that TGTP fits wellwith Rdrp (SI Appendix, Fig. S3). TGTP was therefore synthe-sized (SI Appendix, Fig. S6) and tested as an inhibitor of Rdrpin vitro; unfortunately, there was no inhibition against Rdrp. Inprevious reports, thioguanine was found to act as a slow-binding,reversible, and competitive PLpro inhibitor of SARS-CoV (21)and SARS-CoV-2 (22), indicating further study was necessary to

understand the mechanism of thioguanine and thioguanine analogsagainst SARS-CoV-2. In addition, several known viral polymeraseinhibitors—including acyclovir, famciclovir, penciclovir, ribavirin,cidofovir, and entecavir—and reverse-transcriptase inhibitors—including zalcitabine, nevirapine, efavirenz, abacavir sulfate,tenofovir disoproxil fumarate, adefovir dipivoxyl, delavirdine, andtelbivudine—were also screened in the cell-based assay but noactive compound was identified.

Spike Protein Mutation and Glycosidase Inhibitors. Analysis of the196,276 sequences of S protein variants obtained from theGISAID (Global Initiative on Sharing Avian Influenza Data)database (23) on November 15, 2020 revealed that the S proteinhas already mutated on 1,141 sites of the 1,273 amino acid se-quences (i.e., ∼90% of the S protein has been mutated) (Fig. 5).In addition, the trimeric S protein is highly glycosylated with22 N-glycosites and 2 O-glycosites per monomer (9, 24); the2 O-glycosites and 2 N-glycosites are located in the receptorbinding domain (RBD) region, while some other N-glycosites arefully conserved (e.g., N122, N343). However, glycosylation iscell-specific and the role of glycosylation in S protein during viralinfection and interaction with other host proteins in airway ep-ithelial cells is not well understood and could be a subject forfurther investigation. Nevertheless, despite the broad scope ofmutation in the S protein, it appears that most variants still in-teract with ACE2, suggesting that the RBD domain may havelittle change in conformation and ACE2 could serve as a decoyto block the entry of the virus.In any event, disruption of the glycosylation process of S

protein by α-glucosidases I and II could lead to the formation ofincorrect glycoform, causing protein misfolding, ER-associateddegradation, interference of protein–protein interaction, andinhibition of virus maturation (25). α-Glucosidase was a prom-ising anti–SARS-CoV-2 target since the production of endo-α-1,2-mannosidase (MANEA) was not detectable in humanpneumocytes (26), and a deep mutational scanning studyrevealed that the N343 glycosylation was necessary for the RBDexpression (27). In addition, a computer simulation of S proteinalso suggested the importance of N-glycans at N165 and N234 inmodulating the receptor binding of S protein (28). However,among the five α-glucosidase inhibitors (miglitol, voglibose,acarbose, N-methyl-1-deoxynojirimycin, daucosterol) in our li-brary and several iminocyclitols from previous studies (SI Ap-pendix, Fig. S4) evaluated, none was able to efficiently block theviral infection in the cell-based assay. This preliminary resultseems to suggest that the role of glycosylation in the cell-basedinfection process is not significant, although the assay may notrepresent the complete picture of the virus infection cycle, or thelack of activity may be due to the compensating effect ofMANEA in Vero E6 cells after inhibition of glucosidase (29), or

Fig. 3. Structures of protease inhibitors and their IC50 and Ki values for 3CLprotease inhibition. The values were determined from three independentexperiments using FRET-based enzymatic assays.

Fig. 4. Computer simulation of (A) nelfinavir and (B) boceprevir binding to SARS-CoV-2 3CL protease (PDB ID code 6LU7). Pink dashed lines indicated in-teraction between compounds and protein.

4 of 8 | PNAS Jan et al.https://doi.org/10.1073/pnas.2021579118 Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-

2 infection

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

due to the different natures of cells and virus used in theassay (30).

Inhibitors of SARS-CoV-2 Entry, Nuclear Import, and Replication.Cepharanthine (4, IC50 of 2.8 μM) is an alkaloid isolated fromStephania, exhibiting multiple biological efficacies and rare ad-verse events. It was used for leukopenia, snake bites, and alo-pecia, and was reported to have antiviral activities with multiplemechanisms in laboratory research (31). Recently, it was foundto be able to block virus entry and show a synergistic antiviralactivity with nelfinavir (32). (+)-Emetine hydrochloride (5, IC50of 0.00040 μM) is an antiprotozoan agent approved for amoe-biasis. Emetine was found to be a broad-spectrum inhibitor ofother coronaviruses (33), and it was also identified in this screenas an effective inhibitor of SARS-CoV-2 replication. The car-diotoxicity of emetine could be a concern, and whether it couldbe used at low dose or in combination with other antiviral agents,such as remdesivir, to achieve better clinical benefits remained tobe investigated (34).

Ivermectin (6, IC50 of 4.1 μM) and moxidectin (7, IC50 of 3.1 μM)share the same scaffold and are potent avermectin-type compoundsthat have been used as antiparasitic agents. Their mode of actionis through the enhancement of glutamate-gated chloride chan-nel, leading to chloride influx and paralyzing the parasite. One ofthe antiviral mechanisms was related to the inhibition of proteinsassociated with viral nuclear import (35). However, a studypredicted with a population pharmacokinetic model indicated thatthe concentration of ivermectin in the lung might not achieve thedesired level with even 10 times of approved dosing regimen (36).Both ivermectin and moxidectin exhibited high cytotoxicity in thecell-based assay.Mefloquine (8) belongs to the class of quinine-type antima-

larial agents, and is effective against SARS-CoV-2 infection withIC50 of 3.2 μM. During the pandemic, the antimalarial agentschloroquine and hydroxychloroquine (also used for the treatmentof autoimmune diseases) were reported to exhibit promisinganti–SARS-CoV-2 effect in an early in vitro study (37), leading tolarge clinical trials to evaluate the benefit of chloroquine and

Fig. 5. SARS-CoV-2 spike protein sequence mutation analysis. (A) Analysis of spike protein mutation from the 196,276 sequences revealed 1,141 sites ofmutation in the 1,273 amino acids of spike protein. The spike protein sequence of hCoV-19/Taiwan/4/2020 used in this study was identical to the original virusstrain (UniProt Entry: P0DTC2). Single bottom line: S1 region (residues 13 to 685); double bottom line: S2 region (residues 686 to 1273); yellow bottom line:receptor binding domain (residues 319 to 541); green top line: N-glycosylation motifs; blue top line: O-glycosylation sites; pink: mutation residues; sequencerepresentative: P0DTC2 (UniProt Entry). (B) Top and side view of the S protein indicating residue variants (from strictly conserved to highly variable: white tored), N-glycans (green stick), O-glycosylation sites (blue) in three-dimensional structure (PDB ID code 7CN9).

Jan et al. PNAS | 5 of 8Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-2infection

https://doi.org/10.1073/pnas.2021579118

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

hydroxychloroquine in patients. Although quinine-type compoundsseemed to have some positive activities against SARS-CoV-2 infec-tion, the cardiac side effects of chloroquine and hydroxychloroquinein clinical studies were reported (6). Thus quinine-type compoundswith lower cardiac toxicity and cytotoxicity, such as mefloquine, couldbe explored as an alternative candidate.

Ion Channel Modulator and Ionophores. Four compounds (ivacaftor9, IC50 = 3.7 μM; azelnidipine 10, IC50 = 5.3 μM; penfluridol 11,IC50 = 2.4 μM; dronedarone 12, IC50 = 4.5 μM) identified in ourlibrary with anti–SARS-CoV-2 activities were categorized intoion channel modulators. Electrolyte homeostasis is an importantfactor for viruses to replicate and survive, and many of themexpressed viroporins to control host ion balance (38). Since theion channel displays various roles in the viral life cycle, such asvirus entry and replication, channel modulators are a new type ofbroad-spectrum antiviral agents (39). However, their potentialtoxicity remained to be concerned.Three compounds used as animal-ionophoric antibiotics to

prevent coccidiosis of chicken were identified in this study to beeffective against SARS-CoV-2. Monensin (14) exhibited an IC50of 6.4 μM and maduramicin ammonium (15) has an IC50 of1.3 μM, but both of them showed no selectivity between theantiviral effect and cytotoxicity. However, salinomycin (13)exhibited an excellent antiviral activity (IC50 of 0.00048 μM) andselectivity (CC50 of 13.1 μM, selectivity index > 100). The modeof action for salinomycin is still unclear, but it was proposed todisrupt the endosomal acidification and to block the entry ofviruses into cell and enhance host-directed antiviral responses(40). Autophagy was recently identified to play a role in hostantiviral immunity, and was effective in the inhibition of coro-naviruses (41). Ionophoric compounds were reported to triggerautophagy through inducing an electrolyte imbalance in potassiumand sodium (42), and inhibition of E3-ligase S-phase kinase-associated protein 2 (SKP2) (41). Since hypokalemia and hypo-natremia were common in severe SARS-CoV-2 patients (43), thetherapeutic window of salinomycin and its combination with anelectrolyte supplement, such as Zn-gluconate, could be furtherevaluated.

Traditional Chinese Herbal Medicines. Some well-known traditionalChinese herbal medicines were also tested in the cell-based assay(Fig. 6 and SI Appendix, Table S2). The medical herbs (1.0 g)were extracted by water (20 mL) at room temperature and di-luted with the assay buffer. We found that the aqueous extractsof herbs from Lamiaceae (Perilla frutescens), Mentheae (Menthahaplocalyx), Asteraceae (Taraxacum mongolicum, Tussilago farfara,Chrysanthemum morifolium), Theaceae (Camellia sinensis), Lam-iaceae (Prunella vulgaris, Ocimum basilicum, Salvia hispanica, Nepetatenuifolia, Salvia rosmarinus), Fabaceae (Arachis hypogaea, Spatho-lobus suberectus), and Sapindaceae (Dimocarpus longan, Litchi chi-nensis) families were able to reduce the CPE of SARS-CoV-2 inVero E6 cells when the extracts were diluted to 16- to 960-fold(Fig. 6). These herbs contain flavonoids (myricetin from litchi chi-nensis seed) (44), flavan-3-ol (catechin and epigallocatechin gallatefrom tea and Spatholobus root) (45, 46), caffeic acid derivatives[caftaric acid and chlorogenic acid (47) from purple coneflower orhoneysuckle flower bud or chia seed (48)] (SI Appendix, Fig. S2), andwhether they are related to the antiviral activity remains to be in-vestigated. It was reported that monoterpenes (1, 8-cineole andcamphor from basil leaves) (49), diterpenes [carnosic acid (50) andpatchouli alcohol (51) from rosemary, and patchouli], and triterpenes(ursolic acid from prunella spike and basil) were able to block virusentry and replication. However, the exact mechanisms of theseChinese herbal medicines to inhibit the infection of SARS-CoV-2are still unknown, and synergistic effects of multiple active compo-nents may exist (52).

On the other hand, several fractions of L-fucose–containingpolysaccharides previously isolated and characterized fromGanoderma lucidum (Reishi) were tested in the cell-basedanti–SARS-CoV-2 assay, and the Reishi L-fucose–containingpolysaccharides fraction 3 (RF3) was found to exhibit out-standing antiviral efficacy (2 μg/mL), and it was still active at1,280-fold dilution, with no cytotoxicity (Fig. 6) (53). Althoughthe preliminary results from cell-based experiments cannot bedirectly extrapolated to clinical outcomes, the potential of RF3as anti–SARS-CoV-2 agent is worth of further evaluation.

In Vivo Anti–SARS-CoV-2 Assay. Four of 15 active drugs identifiedfrom the cell-based assay—mefloquine, nelfinavir, salinomycin,and thioguanine—were selected to evaluate their antiinfectiveefficacy in female golden Syrian hamsters. Among the activeherbal medicines, three extracts—RF3, P. frutescens, and M.haplocalyx—were also selected for further evaluation since RF3exhibited a significant antiinfective/cytotoxic selectivity, and P.frutescens and M. haplocalyx were major active gradients inRespireAid TM (NRICM101). Salinomycin was shown to causehigher weight loss than the control group (water administrated),while all other drugs and extracts showed insignificant weightloss. Therefore, the issue of acute toxicity was not a concern inthe chosen drugs and extracts, except for salinomycin (Fig. 7B).In the animal study, hamsters were infected with SARS-CoV-2intranasally at day 0, and after 3-d treatment of orally adminis-tered drugs (at a dose of 30 mg/kg/d) and extracts (200 mg/kg/d)the hamsters were sacrificed and the lungs were collected tomeasure the viral load. Surprisingly, the two compounds, thio-guanine and salinomycin, with better cell-based assay activitiesshowed no significant antiinfective effect in the animal study.However, mefloquine and extract of M. haplocalyx significantlyreduced viral load than that of control (P = 0.005) (Fig. 7A), andnelfinavir, extracts of RF3, and P. frutescens also showed goodantiviral effects (P = 0.03 v.s. control) (Fig. 7A). Overall, in thein vivo assay, mefloquine and nelfinavir were identified as po-tential drug-repurposing agents and extracts of M. haplocalyx, P.frutescens, and RF3 showed potential as anti–SARS-CoV-2herbal candidates.The inconsistency of the antiviral efficacy in cell and animal

assays may originate from the high hydrophilicity and low oralbioavailability of compounds, or from the different infectionmechanisms in the assay models, or the different immune

Fig. 6. Evaluations of antiviral effect of Chinese herbal medicines in serialdilutions were presented as log2(dilution fold). Anti–SARS-CoV-2 infectioneffects of selected Chinese herbal medicines as water extracts (1.0 g/20 mLH2O) and RF3 dissolved in H2O (0.25 mg/mL) are presented. The tested resultsof all Chinese herbal medicines are summarized in SI Appendix, Fig. S7.

6 of 8 | PNAS Jan et al.https://doi.org/10.1073/pnas.2021579118 Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-

2 infection

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

systems of animals. Golden Syrian hamsters were shown to bebetter than mice as animal models for SARS-CoV-2 infectionsince the ACE2 transgenic mice were not readily available (54),and the hamster model was shown to have similar histopatho-logical properties to humans in upper and lower respiratory tractinfections and coherent inflammatory cytokines profiles (55).However, viral clearance was observed in hamsters after day 6postinfection, indicating that hamsters may only mimic the mildhuman COVID-19 case (56). A mouse-adapted SARS-CoV-2strain was reported and demonstrated to represent more severeinfection, but the preparation was time-consuming and the mu-tation in the receptor-binding domain might alter the function ofspike protein (57). Rhesus macaque seems to be a good model forthe development of anti–SARS-CoV-2 agents, as it contains similaranatomy, physiology, and immune systems to that of humans.However, this model is limited by the availability and cost, and itcannot represent the severe case in humans (58). Therefore,developing an appropriate and accessible animal model to mimicthe complete human infection process is an urgent need foraccelerating the development of anti–SARS-CoV-2 agents.

ConclusionIn summary, 15 chemical entities from a library of 2,855 com-pounds approved for human or animal use have been identifiedin this Vero E6 cell-based study to have the anti–SARS-CoV-2activity. These compounds were categorized into five groups asviral protease inhibitors (nelfinavir, boceprevir), guanine analog(thioguanine), inhibitors of SARS-CoV-2 function (cepharanthine,emetine, ivermectin, moxidectin, mefloquine), ion channel modula-tors (ivacaftor, azelnidipine, penfluridol, dronedarone), and iono-phoric antibiotics (salinomycin, monensin, maduramicin). Since thesafety and pharmacological characteristics of these drugs were ex-tensively studied, the preclinical and clinical assessments of the active

compounds identified in this study are expected to be rapid, and canefficiently reduce the time and cost for further development. In ad-dition, several extracts of Chinese herbal medicines and supplementsshowed promising anti–SARS-CoV-2 effects in Vero E6 cell-basedassays, and of particular significance are the species of Asteraceae,Theaceae, Mentheae, and Lamiaceae family, as well as the RF3fraction. A recent study showed that heparan sulfate acted as a cor-eceptor for the S protein (59). However, several heparan sulfate-related structures were shown to be inactive in our cell-based assay,although RF3 exhibited a significant inhibition activity, and its modeof action remains to be investigated. These herbal medicines withanti–SARS-CoV-2 activities could also be interesting sources for thediscovery of new chemical entities as inhibitors of the virus. However,due to the lack of precise animal model for further evaluation ofpromising candidates identified from the cell-based assays, the activeagents reported in this study may have to be further assessed when abetter animal model is available.

Materials and MethodsDetailed information on the compound sources, primary screening, IC50 ofanti–SARS-CoV-2 effect, cytotoxicity in Vero E6 cells, inhibitory activityagainst SARS-CoV-2 protease and Rdrp, computer modeling of inhibitorsbinding to SARS-CoV-2 3CL protease and Rdrp, analysis of S protein variants,and animal study are described in SI Appendix, Materials and Methods.

All experiments involving live SARS-CoV-2 were performed in an animalBSL-3 facility at the Genomics Research Center, Academia Sinica. The studyprotocol was approved by the Institutional Animal Care and Use Committee.

Data Availability. All study data are included in the article and SI Appendix.

ACKNOWLEDGMENTS. We thank the Ministry of Science Technology (MOST108-3114-Y-001-002, MOST 109-0210-01-18-02) and Academia Sinica (AS-SUMMIT-109, AS-KPQ-109-BioMed) for their support.

1. C. Del Rio, P. N. Malani, COVID-19—New insights on a rapidly changing epidemic.

JAMA 323, 1339–1340 (2020).2. Y. A. Helmy et al., The COVID-19 pandemic: A comprehensive review of taxonomy,

genetics, epidemiology, diagnosis, treatment, and control. J. Clin. Med. 9, 1225 (2020).3. M. Hoffmann et al., SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is

blocked by a clinically proven protease inhibitor. Cell 181, 271–280.e8 (2020).4. I. Sola, F. Almazán, S. Zúñiga, L. Enjuanes, Continuous and discontinuous RNA syn-

thesis in coronaviruses. Annu. Rev. Virol. 2, 265–288 (2015).5. M. Wang et al., Remdesivir and chloroquine effectively inhibit the recently emerged

novel coronavirus (2019-nCoV) in vitro. Cell Res. 30, 269–271 (2020).6. O. Mitjà et al.; BCN PEP-CoV-2 RESEARCH GROUP, Hydroxychloroquine for early

treatment of adults with mild Covid-19: A randomized-controlled trial. Clin. Infect.

Dis., ciaa1009 (2020).7. D. E. Gordon et al., A SARS-CoV-2 protein interaction map reveals targets for drug

repurposing. Nature 583, 459–468 (2020).8. M. Bouhaddou et al., The global phosphorylation landscape of SARS-CoV-2 infection.

Cell 182, 685–712.e19 (2020).

9. Y. Watanabe, J. D. Allen, D. Wrapp, J. S. McLellan, M. Crispin, Site-specific glycan

analysis of the SARS-CoV-2 spike. Science 369, 330–333 (2020).10. J. Shang et al., Structural basis of receptor recognition by SARS-CoV-2. Nature 581,

221–224 (2020).11. A. C. Walls et al., Structure, function, and antigenicity of the SARS-CoV-2 spike gly-

coprotein. Cell 181, 281–292.e6 (2020).12. C. Y. Wu et al., Small molecules targeting severe acute respiratory syndrome human

coronavirus. Proc. Natl. Acad. Sci. U.S.A. 101, 10012–10017 (2004).13. S. Matsuyama et al., Enhanced isolation of SARS-CoV-2 by TMPRSS2-expressing cells.

Proc. Natl. Acad. Sci. U.S.A. 117, 7001–7003 (2020).14. N. Yamamoto et al., HIV protease inhibitor nelfinavir inhibits replication of SARS-

associated coronavirus. Biochem. Biophys. Res. Commun. 318, 719–725 (2004).15. B. Cao et al., A trial of lopinavir-ritonavir in adults hospitalized with severe Covid-19.

N. Engl. J. Med. 382, 1787–1799 (2020).16. R. C. Group; RECOVERY Collaborative Group, Lopinavir-ritonavir in patients admitted

to hospital with COVID-19 (RECOVERY): A randomised, controlled, open-label, plat-

form trial. Lancet, S0140-6736(20)32013-4 (2020).

Fig. 7. In vivo anti–SARS-CoV-2 assay conducted in female golden Syrian hamsters. (A) Virus elimination effect of drugs and extracts. Hamsters were infectedwith SAR-CoV-2 by intranasal instillation at day 0, and treated with drugs and extracts orally twice a day (30 mg/kg/d for drugs and 200 mg/kg/d for extracts)continuously for 3 d. After 3 d, the lungs were collected to measure the virus load (n = 5), *P < 0.05; **P < 0.005. (B) Body-weight change after 3-d treatment,n = 5 for test group and n = 6 for the control group.

Jan et al. PNAS | 7 of 8Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-2infection

https://doi.org/10.1073/pnas.2021579118

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

17. U. Arshad et al., Prioritization of anti-SARS-Cov-2 drug repurposing opportunitiesbased on plasma and target site concentrations derived from their established humanpharmacokinetics. Clin. Pharmacol. Ther. 108, 775–790 (2020).

18. Y. M. Shao et al., Structure-based design and synthesis of highly potent SARS-CoV 3CLprotease inhibitors. ChemBioChem 8, 1654–1657 (2007).

19. F. Musarrat et al., The anti-HIV drug nelfinavir mesylate (Viracept) is a potent in-hibitor of cell fusion caused by the SARSCoV-2 spike (S) glycoprotein warrantingfurther evaluation as an antiviral against COVID-19 infections. J. Med. Virol. 92,2087–2095 (2020).

20. N. K. de Boer, A. A. van Bodegraven, B. Jharap, P. de Graaf, C. J. Mulder, Drug insight:Pharmacology and toxicity of thiopurine therapy in patients with IBD. Nat. Clin. Pract.Gastroenterol. Hepatol. 4, 686–694 (2007).

21. C. Y. Chou et al., Thiopurine analogues inhibit papain-like protease of severe acuterespiratory syndrome coronavirus. Biochem. Pharmacol. 75, 1601–1609 (2008).

22. C. D. Swaim et al, 6-Thioguanine blocks SARS-CoV-2 replication by inhibition of PLproprotease activities. bioRxiv [Preprint] (2020) https://doi.org/10.1101/2020.07.01.183020(Accessed 1 July 2020).

23. Y. Shu, J. McCauley, GISAID: Global initiative on sharing all influenza data—Fromvision to reality. Euro Surveill. 22, 2–4 (2017).

24. A. Shajahan, N. T. Supekar, A. S. Gleinich, P. Azadi, Deducing the N- and O- glyco-sylation profile of the spike protein of novel coronavirus SARS-CoV-2. Glycobiology,cwaa042 (2020).

25. S. J. Williams, E. D. Goddard-Borger, α-glucosidase inhibitors as host-directed antiviralagents with potential for the treatment of COVID-19. Biochem. Soc. Trans. 48,1287–1295 (2020).

26. M. Uhlén et al., Proteomics. Tissue-based map of the human proteome. Science 347,1260419 (2015).

27. T. N. Starr et al., Deep mutational scanning of SARS-CoV-2 receptor binding domainreveals constraints on folding and ACE2 binding. Cell 182, 1295–1310.e20 (2020).

28. L. Casalino et al., Beyond shielding: The roles of glycans in the SARS-CoV-2 spikeprotein. ACS Cent. Sci. 6, 1722–1734 (2020).

29. S. E. Moore, R. G. Spiro, Demonstration that Golgi endo-alpha-D-mannosidase pro-vides a glucosidase-independent pathway for the formation of complex N-linkedoligosaccharides of glycoproteins. J. Biol. Chem. 265, 13104–13112 (1990).

30. S. Rajsekharan et al, Repurposing of miglustat to inhibit the coronavirus severe ac-quired respiratory syndrome SARS-CoV-2. bioRxiv [Preprint] (2020) https://doi.org/10.1101/2020.05.18.101691 (Accessed 20 May 2020).

31. C. Bailly, Cepharanthine: An update of its mode of action, pharmacological propertiesand medical applications. Phytomedicine 62, 152956 (2019).

32. H. Ohashi et al, Multidrug treatment with nelfinavir and cepharanthine againstCOVID-19. bioRxiv [Preprint] (2020) https://doi.org/10.1101/2020.04.14.039925 (Ac-cessed 14 April 2020).

33. L. Shen et al., High-throughput screening and identification of potent broad-spectrum inhibitors of coronaviruses. J. Virol. 93, e00023-19 (2019).

34. K. T. Choy et al., Remdesivir, lopinavir, emetine, and homoharringtonine inhibitSARS-CoV-2 replication in vitro. Antiviral Res. 178, 104786 (2020).

35. K. M. Wagstaff, H. Sivakumaran, S. M. Heaton, D. Harrich, D. A. Jans, Ivermectin is aspecific inhibitor of importin α/β-mediated nuclear import able to inhibit replicationof HIV-1 and dengue virus. Biochem. J. 443, 851–856 (2012).

36. V. D. Schmith, J. J. Zhou, L. R. L. Lohmer, The approved dose of ivermectin alone is notthe ideal dose for the treatment of COVID-19. Clin. Pharmacol. Ther. 108, 762–765(2020).

37. J. Liu et al., Hydroxychloroquine, a less toxic derivative of chloroquine, is effective ininhibiting SARS-CoV-2 infection in vitro. Cell Discov. 6, 16 (2020).

38. C. W. Sze, Y. J. Tan, Viral membrane channels: Role and function in the virus life cycle.Viruses 7, 3261–3284 (2015).

39. S. Hover, B. Foster, J. N. Barr, J. Mankouri, Viral dependence on cellular ionchannels—An emerging anti-viral target? J. Gen. Virol. 98, 345–351 (2017).

40. S. Jeon et al., Identification of antiviral drug candidates against SARS-CoV-2 fromFDA-approved drugs. Antimicrob. Agents Chemother. 64, e00819–e00820 (2020).

41. N. C. P. Gassen et al, Analysis of SARS-CoV-2-controlled autophagy reveals spermidine,MK-2206, and niclosamide as putative antiviral therapeutics. bioRxiv[Preprint] (2020)https://doi.org/10.1101/2020.04.15.997254 (Accessed 16 April 2020).

42. A. Kondratskyi, K. Kondratska, R. Skryma, D. J. Klionsky, N. Prevarskaya, Ion channelsin the regulation of autophagy. Autophagy 14, 3–21 (2018).

43. G. Lippi, A. M. South, B. M. Henry, Electrolyte imbalances in patients with severecoronavirus disease 2019 (COVID-19). Ann. Clin. Biochem. 57, 262–265 (2020).

44. M.-S. Yu et al., Identification of myricetin and scutellarein as novel chemical inhibitorsof the SARS coronavirus helicase, nsP13. Bioorg. Med. Chem. Lett. 22, 4049–4054(2012).

45. J. Joseph et al, Green tea and Spirulina extracts inhibit SARS, MERS, and SARS-2 spikepseudotyped virus entry in vitro. bioRxiv [Preprint] (2020) https://doi.org/10.1101/2020.06.20.162701 (Accessed 20 June 2020).

46. M. Menegazzi et al., Protective effect of epigallocatechin-3-gallate (EGCG) in diseaseswith uncontrolled immune activation: Could such a scenario Be helpful to counteractCOVID-19? Int. J. Mol. Sci. 21, E5171 (2020).

47. F. Chen et al., In vitro susceptibility of 10 clinical isolates of SARS coronavirus to se-lected antiviral compounds. J. Clin. Virol. 31, 69–75 (2004).

48. B. Kulczy�nski, J. Kobus-Cisowska, M. Taczanowski, D. Kmiecik, A. Gramza-Micha-łowska, The chemical composition and nutritional value of chia seeds-current state ofknowledge. Nutrients 11, 1242 (2019).

49. T. F. Kubiça, S. H. Alves, R. Weiblen, L. T. Lovato, In vitro inhibition of the bovine viraldiarrhoea virus by the essential oil of Ocimum basilicum (basil) and monoterpenes.Braz. J. Microbiol. 45, 209–214 (2014).

50. S. Birti�c, P. Dussort, F. X. Pierre, A. C. Bily, M. Roller, Carnosic acid. Phytochemistry 115,9–19 (2015).

51. H. Kiyohara et al., Patchouli alcohol: In vitro direct anti-influenza virus sesquiterpenein Pogostemon cablin Benth. J. Nat. Med. 66, 55–61 (2012).

52. J. S. Mani et al., Natural product-derived phytochemicals as potential agents againstcoronaviruses: A review. Virus Res. 284, 197989 (2020).

53. S. F. Liao et al., Immunization of fucose-containing polysaccharides from Reishimushroom induces antibodies to tumor-associated Globo H-series epitopes. Proc.Natl. Acad. Sci. U.S.A. 110, 13809–13814 (2013).

54. P. V’Kovski, A. Kratzel, S. Steiner, H. Stalder, V. Thiel, Coronavirus biology and rep-lication: Implications for SARS-CoV-2. Nat. Rev. Microbiol., 10.1038/s41579-020-00468-6 (2020).

55. S. F. Sia et al., Pathogenesis and transmission of SARS-CoV-2 in golden hamsters.Nature 583, 834–838 (2020).

56. M. Imai et al., Syrian hamsters as a small animal model for SARS-CoV-2 infection andcountermeasure development. Proc. Natl. Acad. Sci. U.S.A. 117, 16587–16595 (2020).

57. K. H. Dinnon, 3rd et al., A mouse-adapted model of SARS-CoV-2 to test COVID-19countermeasures. Nature 586, 560–566 (2020).

58. V. J. Munster et al., Respiratory disease in rhesus macaques inoculated withSARS-CoV-2. Nature 585, 268–272 (2020).

59. T. M. Clausen et al., SARS-CoV-2 infection depends on cellular heparan sulfate andACE2. Cell 183, 1043–1057.e15 (2020).

8 of 8 | PNAS Jan et al.https://doi.org/10.1073/pnas.2021579118 Identification of existing pharmaceuticals and herbal medicines as inhibitors of SARS-CoV-

2 infection

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021